One source of truth for how we prove what Hakkoda and IBM build.
We are building an alternative to one-off, deadline-driven collateral: a structured library of client proof, mapped to capabilities and verticals, that compounds in quality and reuse with every engagement. The output is sales and marketing material that is faster to produce, consistent across teams, and defensible in front of any buyer, because every claim traces to a verified source.
157
Client proof points cataloged and structured
8
Capability areas defined and proof-mapped
13
Industry verticals covered
~98
Records with confirmed, quantified outcomes
The data layer
Every record is extracted from raw source material, win-story decks, the master case-study deck, engineer write-ups, and the public site, then normalized into one structured schema: capability, vertical, delivering team, architecture, metric, and proof status. Unstructured use cases become queryable, comparable data.
It is a living source of truth that grows with every new engagement, and it feeds both off-the-shelf assets and custom collateral built on the fly for a specific client or pursuit.
Today it lives as a structured markdown library. Next it moves into a database that powers an interactive, searchable internal asset library and data room for sales and marketing.
The system, end to end
Intake · new client use cases and builds, logged at close-out
↓
Layer 1 · Data
Source of truth
157 structured records, growing with every engagement
13 verticals8 capabilitiesDelivery + proof statusNamed metrics and quotes
↓
Layer 2 · Logic
Intelligence
Turns records into the right proof, and spots what is missing
Locked Carbon and IBM-compliant templates with surgical functionality
Carbon / IBM brandingShare CTAsArchitecture diagramsDemo videosProof tables
↓
Output · three-pagers, leadership briefings, internal data room
01Where we are
The V1 source of truth is builtLive
A first version is in place: 157 client engagements captured in one schema, capability, vertical, delivering team, architecture, metric, and proof status. The top three verticals (Healthcare 42, Financial Services 23, Retail & CPG 22) are already deep enough to staff most enterprise conversations, and the library grows with every engagement.
A matching engine connects proof to capabilityLive
A logic layer ranks the strongest proof for any capability and separates Hakkoda-delivered from IBM-delivered work, so the right evidence surfaces in seconds instead of a manual hunt.
A project intake web appLive
A Carbon-styled intake form is deployed and capturing new client builds at engagement close-out, writing each submission straight into the source-of-truth pipeline. Growing the library is now a standing workflow, not a manual scramble before each pursuit.
Capability three-pagers in draft5 drafts
Intelligent Automation, Decision Intelligence, Data Modernization, Customer & Product Intelligence, and Agentic Orchestration are drafted on a locked, co-branded template with supporting architecture diagrams.
First partner three-pagerSnowflake
Snowflake is the first partner three-pager, generated from the source of truth on the same locked template and already deployed for review.
First vertical three-pagerHealthcare
Healthcare is the first industry cut. The same engine pulls every Hakkoda and IBM build in the vertical across all capabilities, our deepest dataset at 42 records and 37 confirmed, with each proof row tagged to the capability it demonstrates.
Once this version of the source of truth was standing, the Snowflake three-pager above was generated from one prompt. The structured proof did the work: the right clients, metrics, and quotes were already queryable, so the asset assembled itself on the locked template.
"Let's test our capabilities and spin up a three-page Snowflake x Hakkoda / IBM 3-pager, using the Agentic Orchestration template as is for the presentation layer and structure."
Each engagement is extracted into one structured record with the same fifteen fields, so unstructured use cases become comparable, queryable data. Four of those fields are search facets: stack them and 157 records collapse to the exact proof a pursuit needs.
filterable facetrequired / conditional shown at right
Confirmed 98Structural 53Directional 12UnconfirmedMethodology only
Architecture and tech stack is keyword-searchable (Snowflake, watsonx, dbt), and a present quote or diagram acts as a flag.
Example: stack three facets
Data Modernization+Healthcare+Confirmed→focused result set
Beth Israel Lahey, Edwards Lifesciences, City of Hope, CMS, Advocate Health and peers, each a named build with a confirmed metric, ready to drop into an asset.
Today these facets filter the structured markdown library. In Phase 2 the same model powers an interactive, searchable database and the internal data room.
03How a capability becomes a finished asset
Picking a capability runs the library through one repeatable logic pass. It selects the right proof, separates who delivered it, ranks it by strength, and returns a built asset where every row traces to a source. The same pass also shows us where proof is thin.
1 · Capability match
Every record is tagged to one or more of the eight capability buckets. Choosing a capability pulls every engagement mapped to it, across all thirteen verticals, in a single query.
2 · Two proof benches, never blended
Results split by who delivered the work. Hakkoda-delivered builds form the spine of the asset; IBM Consulting builds form a deeper bench shown in a separate, labeled group. We never present one team's work as the other's.
3 · Proof discipline, applied automatically
Confirmed records lead with their hard metric. Structural and directional records appear as credible builds with no invented number. Projected figures stay labeled as projections, never as realized results.
4 · Pattern ranking
The strongest named outcomes sort to the top and grade down, with structural and directional proof always settling at the bottom. A reader's eye lands on the best evidence first, and the proof never breaks up mid-table.
5 · Weak proof held back
Unconfirmed or in-progress records are withheld automatically. Nothing reaches a client-facing asset on a claim we cannot stand behind.
6 · Gap detection
The same pass reveals where coverage is thin. Agentic Orchestration, for example, shows strong IBM Consulting proof but no Hakkoda-delivered spine yet, turning the library into a roadmap for what to collect next.
7 · Guardrails on every buildAutomated
A locked, co-branded template and an automated check run on every asset before it ships. Together they reject a fabricated client, a banned phrase, or a mislabeled delivery. If a name is not in the source of truth, it does not render.
8 · Human in the loopLive
New submissions from the intake app are not trusted on arrival. Each is reviewed and validated by a person, attribution, metric, and source checked, before it is promoted into the verified source of truth and becomes eligible for any asset. Automation assembles; a human confirms.
Under the hood
The library compiles into one normalized index, a single de-duplicated record per client across all source files. Three Python scripts run over it: a generator that selects, formats, and ranks proof rows; a linter that blocks fabricated clients and off-brand copy; and a shell-lock check that guarantees only vetted content changes between assets. Output drops into a locked Carbon and IBM-compliant HTML template.
Worked example: a capability cut
Intelligent Automation→7 Hakkoda spine+33 IBM Consulting→ranked, source-traced asset
The engine pulls every Intelligent Automation build, splits Hakkoda-delivered from IBM Consulting, applies proof discipline, and ranks each group metrics-first with structural proof at the base. The asset leads with a Hakkoda spine of named builds (FIS, Under Armour, Conagra), backed by 33 IBM Consulting builds in the expandable group. Decision Intelligence runs the identical pass: a Hakkoda spine plus 19 IBM Consulting builds.
Worked example: a vertical cut
Healthcare→16 Hakkoda spine+11 IBM Consulting→capability-tagged asset
The same engine pivoted to an industry. It pulls every Healthcare build across all capabilities, our deepest vertical at 28 Hakkoda-delivered builds, features the strongest 16, and tags each proof row with the capability it demonstrates. Edwards Lifesciences and CMS lead on metrics; recognizable names like Boston Children's and City of Hope sit below.
04Why the numbers can be trusted
Every record carries a proof status. No metric is presented as a client outcome unless it is confirmed in a source. Delivery is balanced and labeled: roughly 72 records are Hakkoda-delivered and the remainder IBM-delivered (IBM Consulting and IBM teams), with co-branding flagged for review before anything goes client-facing.
~98
Confirmed
Hard, quantified outcomes verified in a case study or engineer source. The backbone of every asset.
~53
Structural
Named build, named client, real architecture, where no percentage metric exists yet. Credible without a number.
~12
Directional
Projected or estimated figures, always labeled as such and never passed off as realized results.
05Roadmap
Now
Polish the five capability drafts and the Snowflake partner draft toward deployable quality
Marketing review next week with Millie's team: Daniel on hosting and delivery, Roberto on brand compliance, Zach on client and data validation
Next
Evolve the first set of three-pagers into deployable, brand-approved assets hosted on Cloudflare
Wire full functionality across them: share CTAs, architecture diagrams, and demo videos
Then
A self-serve, access-gated data room for sales and marketing
Members generate custom assets on the fly from the validated source of truth
Built on locked, compliant templates, CSS, and brand guidelines
Harden the now-live data loop from V1 toward full automation: intake capture is live today; next is streamlined human validation, auto-promotion into the source of truth, and recompiling the index as new builds land, moving toward richer capture where teams upload files, drawings, media, recorded demos, and other asset ingredients with each new build